Multiple Regression Analysis Theory

Properties from Multiple Regression using Matrices

Property 1:
Coefficient matrix regression

Proof: Define the n × 1 vector E such that E = Y − XB. The goal of the least-squares method is to find the (k+1) × 1 vector such that ETE is minimized. Now note that

image9126

image9127

This last equality is true since BTXTY  is a scalar and so

image9128

Setting the first derivative of ETE equal to zero, we get

Thus
image9130

and so
image9125

This is essentially another proof of Theorem 1 of Least Squares for Multiple Regression.

Property 2:
image1856

Proof: This follows from Property 1 and Definition 1 of Multiple Regression using Matrices since

image9133

Property A: The hat matrix H is symmetric and idempotent, i.e. HT = H and HH = H. The same is true for I − H.

Proof: First we observe that (XTX)-1 is symmetric since

image9134

Now
image9135

Also
image9136

This shows symmetry. We now show idempotency.

image9137

image9138

Property 3: B is an unbiased estimator of β, i.e. E[B] = β

Proof: By Property 1image9125

But
image9107

and so we have

image9139

image9140

Thusimage9141

since it is assumed that E[B] = β.

Property 4: The covariance matrix of B can be represented by

image9109

Proof: Based on the assumption that var(εi) = σ2 for all i and cov(εi, εi) = 0 for all i ≠ j, it follows that cov(ε) = E[εεT] = σ2I. As we have seen in the proof of Property 3

image9142

Thus by Property 3
image9143

Hence
image9144

image9145

image9146

Properties from Multiple Regression in Excel

Property B: For any n × n matrix A and n × 1 vectors Y and Z

image9147

Proof:

image9148

image9149

Thus

image9150

image9151

Since Tr(E[C]) = E[Tr(C)] for any square matrix C, Tr(CD) = Tr(DC) and b = Tr(b) for any scalar b, we have

image9152

image9153

image9154

Consequently
image9155

and so
image9156

Property C:

image9157

Proof: We first note that by Property 1

image9158

By Property 3 of Multiple Regression Analysis in Excel and the above equality

image9159

image9160

image9161

Property 4: MSRes is an unbiased estimator of σ2, the variance of the error terms εi

Proof: By Property 1 and C

image9162

By Property B

image9163where
image9164Butimage9165

and soimage9166

Nowimage9167

image9168

Since Y = Xβ + ε, it follows that E[Y] = E[] + E[ε] = , and so we have

image9169

image9170

Putting it all together we have

image9171

Finallyimage9172

2 thoughts on “Multiple Regression Analysis Theory”

  1. Hi Charles,
    Your site provides someone with only limited understanding great insight into statistical methods. Many thanks.
    Please could you explain how you differentiate the matrices in Property 1?

    Reply

Leave a Comment